Synchronized refixation saccades in enhanced VVOR test. A new application for PR score

1 Introduction

The compensatory saccadic response (CSR) that occurs after a head thrust or impulse is a sign that at bedside easily identifies and locates the impaired vestibular function and deficient vestibulo-ocular reflex (VOR) [7]. It is radically different from the ocular movement commanded by VOR in the head impulse test (HIT). The normal VOR “slow phase” is a short time latency (8ms) ocular movement that mimics the head movement magnitude because its neural pathway is a continuous discharging neural network of three neurons integrated in pontine vestibular nuclei [8]. The abnormal CSR has longer time responses (80–100 ms) and magnitude than the head movement because its pathway, mainly located in the paramedian pontine reticular formation, is part of a more complex neural network triggered by an adaptative, volitive or preprogramed central mechanism which acts a response to retinal slip [4, 13].

The recording of the head and eye movement during the HIT, both with the scleral search coil or adapted video, made possible to better characterize the eye response in the normal and vestibular deficient subject; the former system is used more clinically and now the test is also called the video head-impulse test (vHIT) [24]. In terms of timing of the response to the head movement CSR are classified as overt and covert [13]. Attending to vHIT plots of the response in patients with unilateral vestibulopathy, Batuecas et al. [3] described two new patterns: it was called a synchronic CSR pattern when most of the saccadic responses appear at the nearly same latency in the different head impulses performed; and asynchronic CSR pattern when the saccadic responses appeared at a different time lapse in the successive head impulses. Interestingly, the former patients had a significant lower dizziness handicap inventory (DHI) score than the latter ones.

After this clinical observation, the same research group developed a quantification method for these saccadic aggrupation visual patterns defining then the Perez-Rey (PR) score [21] as shown in Fig. 1. This dimensionless number (0–100) has shown to be interesting for the objective evaluation of compensation [1, 23], to follow the dynamic process [3, 17] and to guide vestibular rehabilitation [16].

OPEN IN VIEWER

Fig. 1

PR score for classic vHIT test. On the left side with a low PR score of 8, a left side vestibular neurectomy patient vHIT saccadic responses analysis is plotted. First saccadic responses (x marked saccades) appear at the same time at the end of all the performed head impulses. On the right side, a different right side vestibular neurectomy patient with high PR score of 68. First saccadic responses (x marked saccades) appear at any time during and after the head impulse. As is plotted at the bottom of the figure, PR score goes from 0, maximum gathered saccades to 100 minimum gathered scores (maximum scattered). Orange lines are eye movement plots, red and blue are head right and left respectively plots, cross is marking first appearance saccades, asterisk is marking second appearance saccades and straight cross is marking third appearance saccades.

Since its introduction, the vHIT device has also helped to quantify other vestibular and visual bedside tests [24], in particular visually-enhanced VOR test (VVOR) [20, 22]. This is a clinical test that quantifies VOR gain values at different frequencies of stimulation [20]. Interestingly at higher frequencies (> 1Hz), due to the visual pursuit and optokinetic reflex influence, VOR gain quantification decreases in this test and, a high correlation with vHIT gain values has been shown [22].

Aimed by this correlation between VOR gain values obtained with quantified VVOR and HIT tests the main objectives of this study were 1) to develop a quantification method to obtain a PR score adapted to the VVOR test and 2) to evaluate the correlation of the PR score obtained with quantified VVOR with the PR score of the vHIT test.

3 Results

The PR_VVOR was successfully calculated from all VVOR data. PR score algorithm was implemented and shared in open source terms in GitHub public repository https://github.com/bendermh/VVOR spe-cifically developed PR algorithm for VVOR testing is coded in “prScoreVVR.m” (https://github.com/bendermh/VVOR/blob/master/prScoreVVR.m) MATLAB file.

The validation study of the PR_VVOR was performed with the data of 30 patients: 11 post-neurectomy (due to vestibular schwannoma completely resected) and 19 with subacute vestibular neuritis. Normal eye motility, in particular smooth pursuit, was recorded in all the patients. Demographic data is summarized in Table 1 and age density plots in Fig. 3.

OPEN IN VIEWER

Fig. 3

Density plot showing the age distribution, included pa-tients’ records grouped by clinical diagnose.

Pearson’s correlation coefficient (R²) was for the overall sample 0.92 (p < 0.001) and the 95% confidence interval (CI95) was 0.85 –0.96. In vestibular neurectomy patients, R² was 0.94 (p < 0.001) and CI95 : 0.79 –0.98, in vestibular neuritis patients, R² was 0.90 (p < 0.001) and CI95:0.76–0.96. Linear regression plots are shown in Fig. 4.

OPEN IN VIEWER

Fig. 4

Linear regression for VVOR and vHIT PR score values. In plot A, regression line (blue line) and data dots are plotted, grey area is the 95% confidence interval for regression line. In this figure PR scores from VVOR and VHIT data are plotted for all the patients included in this study (R² = 0.92). In plot B, residual –quantiles plot of linear regression plotted in A is shown. Plots in C are the regression lines between VVOR and vHIT PR scores split in diagnostic groups, vestibular neuritis in left side (R² = 0.90) and vestibular schwannoma with surgical neurectomy in right side (R² = 0.94). Regression lines are blue lines and grey areas are the 95% confidence interval for each regression line. All the obtained R² values where statistically significant (p < 0.001).

In the gain LME model, only the gain of the VOR in the vHIT and VVOR showed statistical influence (p < 0.001); a trend to significant influence was also for lesion side (p = 0.061) in Wald X² test. Nor age, sex, clinical group, time since hypofunction neither presence of spontaneous nystagmus did show statistical association with VOR gain values.

In the PR LME model, only PR_VHIT and PR_VVOR showed statistical association with PR score in Wald X² test (p = 0.008); a trend to significant influence was also found for the clinical group (p = 0.077). Nor age, sex, lesion side, time since hypofunction neither presence of spontaneous nystagmus did show statistical association with PR score values in Wald X² test.

The result in the VVOR test comes from the combination of two findings. In the first instance, the sinusoidal pattern of eye velocity is easily observed while testing. The quality of the response is defined at a glance according to how it mimics or not with the head velocity. It can be analyzed offline and the gain result (the relation between eye and head velocity) has been shown to be different between cases of unilateral and bilateral vestibulopathy whether peripheral or central [19, 20]. A second observable sign is the CSR, which occur whenever there is a significant amount of vestibular deficit and corresponds to a reduction in gain of the VOR. These saccades will head to the unaffected side (as spontaneous nystagmus) mostly when head is moved toward the affected side. They will appear with different shapes according to the methodology of the current system with which it is recorded [20] and they have to be excluded in order to analyze the relation of eye to head velocity as previously mentioned. In the cases that, additionally, a smooth pursuit disorder is part of the disease, both, the pattern of eye velocity and CSR, will be extremely abnormal like in the case of CANVAS [18]. Our paper is the first to analyze the characteristics of those CSR during VVOR testing in patients with unilateral peripheral vestibulopathy: we have found that their coefficient of variation (PR score) strongly correlates with the vHIT one. This PR_VVOR calculation relies in proper methodology. We have followed the indications by Soriano-Reixach et al. [22] described above.

Whatever any of these recommendations are not fulfilled, the results will probably not be comparable. Therefore, own laboratory results are needed to perform; for them it is imperative to consider the head oscillation frequency, its standard deviation and the number of computed head oscillations. There is a relative lower absolute mean value in PR score of the PR_VHIT presented in this work when it is compared with previously published results [6]. For the present research, we have used the original published open-source system [21] to compute PR score, while in the latter case the authors used a commercially available equipment to compute the PR score, that has not been clinically validated. Despite we believe these differences between PR systems are due to the saccadic responses recognition algorithm and also, because of the differences between patients’ sample, we recommend to obtain their own normality data according to the method in use, especially when using non-validated calculation methods.

Despite the strong positive correlation [9] found in this study between both PR scores (R² = 0.92), an important statistical difference was detected in the gain result in vHIT and VVOR test. This is an expected finding given the differences between both tests. The unpredictability and high frequency (5 Hz) of head impulse test ensures that irregular vestibular afferences of VOR [10] are commanding eye responses in vHIT [8]. However, in the VVOR testing, stimuli are at lower frequency (1-2 Hz). During that sinusoidal stimulation not only will stimulate regular vestibular afferences of VOR response [10], but also (due to predictability) smooth pursuit and optokinetic reflex will contribute to the final response, in particular at < 1 Hz head stimulation frequency [13].

Under these physiological evidences, we can expect, that gain and PR scores, will be different between vHIT and VVOR tests but well correlated in each of them: gain of VOR in vHIT and PR_VHIT, and gain of VOR in VVOR and PR_VVOR as shown before [20, 22] and in this study. This suggests that we are measuring the same gain and PR score parameters slightly different but strongly related because of the different physiological conditions between tests and PR score. We can conclude that measurement of saccade synchronicity in both tests is possible (Fig. 5).

OPEN IN VIEWER

Fig. 5

Example of gathered and scattered PR scores from two patients included in this study are plotted in both (simultaneously collected) vHIT and VVOR tests. On the top of the figure, vHIT in left and VVOR at right are plotted both with low (gathered) PR scores of 8 and 12, respectively, from one patient with vestibular neurectomy. As can be observed, vHIT saccades appears always at the same time intervals and VVOR saccades appears also always at the same head movement time for each oscillation, a little after the head velocity peak. On the contrary, at the bottom of the figure a patient with subacute vestibular neuritis is plotted. Saccadic responses with higher PR values in reference to the previous patient appears in both vHIT and VVOR testing (54 and 58 values in PR score respectively). Observe that at this patient in VVOR testing, saccades appear at different moments of head oscillation; even during contralateral head oscillation (see saccadic responses around second 5).

But because of these statistical differences found in LME model between PR_VHIT and PR_VVOR and the limited diagnostic categories included in this study, further studies should expand the PR_VVOR clinical characteristics and correlation with other vestibular markers of vestibular compensation, similar to the findings actually described for the variations and clinical meaning of PR score in vHIT testing [6, 21].

PR score for both vHIT and VVOR clinical tests did not show any differences between other clinical and measuring variables according LME statistical models. Among these variables, it is especially significant that the presence of SP nystagmus was not a statistical influence for both PR score and VOR gain values. This suggests that PR score is an independent parameter not influenced for the presence of spontaneous nystagmus, but it should be considered that our (retrospective) validation study was not specifically designed to study SP nystagmus influence in PR score. This is somewhat controversial in reference to a very recent publication that showed how the intensity of SP nystagmus at diagnosis will influence not only hospital stay in patients with VN, but also short-term prognosis of compensation [11]. We do not think there is true a controversy, as we are not considering the intensity of the response, only the bedside evaluation. Moreover, SP nystagmus is a static sign of a VOR deficit while CSR, in our method of testing, represent a more complex response during dynamic testing of a deficient VOR [12]. And finally, we know, as such, that [15] the patients who presented CSR are partly due to: 1) VOR is reduced so while moving the head towards the affected side, then the eye drifts with the head and a CSR is needed to keep the eye on target, 2) head motion revealed spontaneous nystagmus, and 3) gaze-evoked nystagmus when the head is at its maximal deviated position and a mild 1^st degree spontaneous nystagmus could occur. Upon that, during low frequency testing the action of the optokinetic-reflex (OKR) contributes to the eye stabilization; however, at high frequencies, the vestibulo-ocular reflex overrules the OKR in particular when the scene is stationary. We consider that the best way to analyze the contribution of each should be by analyzing in a longitudinal follow-up study the different manifestation of SP nystagmus and PR scoring.

We conclude that according to the correlation results obtained in this study (R² = 0.92), the new developed PR score for synchronization measurement of saccadic responses in VVOR testing is a valid method which outputs synchronization values that strongly correlates with PR score in vHIT test (see Figs. 4 5).